Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Need advise with measuring USB2 high speed (480MHz) signals - oscilloscope

Status
Not open for further replies.

userx2

Full Member level 3
Joined
Sep 19, 2008
Messages
151
Helped
2
Reputation
4
Reaction score
2
Trophy points
1,298
Activity points
2,887
Hello friends

I am trying to measure high speed USB signals with an oscilloscope.
The signals are differential and run at 480Mb
The impedance of the tracks is 90 Ohm (Zdiff)

Problem:
I have 4 proto boards (10 layer) rev 1 and all working on USB2 high speed 100%

I now have a heap of new new proto boards (same design different board maker) rev 2and none work on USB2 high speed.
They claim the impedance has been matched and sent me the report and blah.

So I need to try and find the problem by measuring something.

I have a suitable scope with 50 Ohm inputs but no probes (they cost a fortune!!)

I tried to made 2 'probes' with RG58 coax and a series resistor of 1k, soldered directly onto the USB conenctor on the pcb. I soldered the shields together and grounded them at the pcb connector.
The probes are length matched as well.

This does measure but I found that the USB communication stops working and gets full of errors (on a working board).
It gets better when I unplug my probes' from the scope inputs but of course I then can't measure.

I upped the series resistors to 1k2 and this helped a little but still not satisfactory.

At 2k2, the scope no longer measures any usable waveforms.

My question, what am I doing wrong? I would have thought that even a 1k series resistor on the probe input should not load the signals enough to start introducing errors.

I have a (borrowed $$$$) 12GHz differential probe and that works great but I cannot see any eye patterns and other abnormalities with that.

Is there anything else I can do (cheaply) to measure these signals without the problems as above?


Apart from getting the boards re-made somewhere else, I am actually out of ideas on how to find this problem.

Interstingly, my USB analyzer sees errors on packets from my pch host side as well as from the device (memory stick) that is conencted to my host.

Best regards
X
 

Homemade resistive probes are always an option for cheap wide band measurements. But it could be that the DC load disturbs USB operation during initial speed negotiation, did you try AC coupling? Or the already marginal signal quality of your design is further worsened by connecting the probe load.

It's unusual to need true impedance controlled manufacturing for HS USB, systematical impedance design should be sufficient, particularly for usual trace length up to e.g. 10 cm. You can refer to the fairly wide cable impedance specification.

Either your board impedance is really far-off, you have a different board signal quality problem like wrong ground layout, cross talk or similar - or it's a completely different problem, e.g. related to the USB interface device.

I'm a bit surprised that you are unable to see typical signal patterns with a active differential probe.
 

Hello, thanks for the reply.

All you points are valid.

AC coupling? Do you mean just a small cap in series with my 1k2 probe?

I used 1k2 metal film resistors. I read that those may be no no good for high frequencies.
I ordered carbon composite ones yesterday but that may not help.


Unfortunately I don't know.

I have designed heaps of high speed boards and everything has of course been painstakingly designed to be correct several 1000s of my boards work just fine - with USB high speed.

I do suspect some sort of manufacturing defect but the manufacturer does not want to know about it as I have no proof (yet). Remaking is several $1000 so I need to be sure before commencing.

All I can do at the moment is try and compare lots of measurements to working boards.


The signals I measure with that ($$$) differential probe acually look ok. The signals form my processor (host) look a lot better than the signals from the device (memory stick).

Yet I see heaps of CRC errors on the USB analyzer. They appear to be coming from the device side though.

Been on this for a whole week already.

Regards
X
 

Status
Not open for further replies.

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top